Submission: GigaPose+GenFlow (RGBD)/T-LESS/threshold = 0.0

Download submission
Submission name threshold = 0.0
Submission time (UTC) Sept. 13, 2024, 11:05 p.m.
User sp9103
Task Model-based 6D detection of unseen objects
Dataset T-LESS
Description
Evaluation scores
AP:0.467
AP_25:-1.000
AP_25_mm:-1.000
AP_MSPD:0.473
AP_MSSD:0.462
AP_MSSD_mm:-1.000
average_time_per_image:5.922

Method: GigaPose+GenFlow (RGBD)

User sp9103
Publication Genflow: Generalizable recurrent flow for 6d pose refinement of novel objects, CVPR 2024
Implementation -
Training image modalities RGB-D
Test image modalities RGB-D
Description

Submitted to: BOP Challenge 2024

Training data: MegaPose-GSO and MegaPose-ShapeNetCore

Used 3D models: Default,

Notes: We use GigaPose [A] to extract 5 hypothesis and run GenFlow[B]. In this submission, CNOS_fastSAM [A] detections are used

[A] Nguyen et al.: GigaPose: Fast and Robust Novel Object Pose Estimation via One Correspondence, CVPR 2024.

[B] Genflow: Generalizable recurrent flow for 6d pose refinement of novel objects, CVPR 2024

Computer specifications RTX 3090 Ti